Recent methodological innovations for integrating randomized controlled trials (RCTs) and non-randomized studies strategically utilize the strengths of each design and provide opportunities to investigate scientific questions that RCTs or non-randomized studies alone cannot answer. Though not previously deployed, applying these new methods to HIV and substance abuse treatment research can reveal the strengths and limitations of evidence from different study designs and provide better insight into the real-world effectiveness of interventions designed to improve individual and public health. This application uses data from a clinical trial of methadone maintenance therapy (MMT) for HIV-infected Malaysian prisoners where some participants were randomized and some were given the choice of MMT or no MMT (NIDA R01DA025943). The intervention focused on within-prison initiation of MMT to improve HIV and substance abuse treatment outcomes after release. We apply recently-developed statistical approaches for integrating randomized and non-randomized studies to evaluate the weaknesses of each design and utilize their strengths to answer new methodological and substantive questions relevant to substance abuse and HIV treatment research. The Specific Aims for this F30 application are to: (1) Evaluate whether a non-randomized study of MMT can produce similar treatment effect estimates as a RCT. We will examine whether commonly-used and cutting-edge statistical methods can produce estimates similar to RCT findings from nonrandomized participant data. (2) Generalize the RCT findings by integrating RCT and non-randomized study data. Non-randomized study participants who selected MMT are more representative than RCT participants of those who would receive MMT if it were offered outside the research setting. Using recently developed methods for combining RCTs and nonrandomized studies, we will adjust RCT participants to reflect characteristics of non-randomized study participants who selected MMT, thus providing more generalizable effectiveness estimates of the impact of MMT among HIV-infected prisoners. (3) Estimate the effects of treatment preference on MMT retention and HIV and substance abuse treatment outcomes. When individuals receive their preferred treatment, they may be more adherent and thus have better outcomes. By extending methodology from the preference trial literature, we will integrate the RCT and non-randomized study data to estimate the effects of treatment preferences on MMT adherence and HIV and substance abuse treatment outcomes. Through this 3-year fellowship, the applicant will gain training in advanced statistical and epidemiologic methods and extensive preparation from a multidisciplinary mentoring team at Yale University for a career as an academic physician-scientist working at the interface of substance abuse and HIV.